Finrl Rl Trading

Use ensemble deep reinforcement learning (A2C, DDPG, PPO, TD3, SAC) to execute automated multi-market stock trading with

Security Scan

This is an instruction-only RL trading skill whose high-level purpose (ensemble DRL trading/backtesting) matches the content, but the runtime instructions and metadata are inconsistent (undeclared runtime requirements, implicit host commands, and missing credential declarations) — review before running, especially for live execution.

Purpose & Capability

The skill claims DRL-based multi-market trading and the SKILL.md contains DRL/backtest/trading pipelines and use cases (Alpaca, ZVT, data sources). That purpose explains needing Python, market data providers, and broker APIs. However the skill manifest declares no required binaries, env vars, or install spec, while SKILL.md explicitly requires Python 3.12+ and the 'uv' package manager and references running Python commands — a mismatch between declared requirements and actual runtime expectations.

Instruction Scope

SKILL.md instructs the agent to run host Python commands (precondition checks like importing zvt, running recorders, creating ~/.zvt test files, and pip install suggestions) and to reload seed.yaml before behavioral decisions. It references environment variables (ZVT_HOME) and invoking data recorders that will access external data providers and potentially brokers. Those runtime actions reach into the host filesystem and network and are not surfaced in the manifest, giving the agent ability to execute arbitrary Python on the host if used as-is.

Install Mechanism

No install spec or code files are included (instruction-only), so nothing will be automatically downloaded or written by an installer. This lowers filesystem/remote-install risk. Still, the SKILL.md expects Python 3.12+ and a 'uv' package manager but does not provide an install recipe — the agent or user would likely run pip/uv commands manually per the instructions.

Credentials

The skill declares no required credentials or env vars, yet the content describes interacting with data providers (eastmoney, joinquant, akshare) and brokers (Alpaca) which normally require API keys/credentials. The SKILL.md also references ZVT_HOME and suggests writing to ~/.zvt. The absence of declared credential requirements is an inconsistency: if you run live or paper trading flows you'll need to supply sensitive API keys, but the skill does not document how it expects to receive or store them.

Persistence & Privilege

Flags show always:false and disable-model-invocation:false (normal). The skill does not request permanent inclusion or to modify other skills. Runtime instructions do ask to read and re-read seed.yaml and write to its own data directories (~/.zvt) which is within its scope; no evidence it modifies system-wide agent settings or other skills.

Scan Findings in Context

[no_code_files] expected: The regex-based scanner had nothing to analyze because this is an instruction-only skill (no code files). That is expected for a documentation/instruction crystal, but absence of code does not imply safety — the SKILL.md itself is the runtime behavior.

What to consider before installing

Before installing or running this skill: (1) Treat it as a set of host instructions — it expects you to run Python and may run Python commands that touch ~/.zvt and call external data/broker APIs. Review seed.yaml and references (LOCKS, ANTI_PATTERNS) to understand fatal constraints. (2) Do not run live/paper trading with real API keys until you understand what commands will be executed; the skill does not declare where/how it will store or use broker credentials. Use sandboxed environment or container and a throwaway account for testing. (3) Ensure your Python version and package manager match SKILL.md (it claims Python 3.12+ and 'uv'); the manifest not listing these is an inconsistency to fix. (4) Verify source and license (homepage/source unknown, LICENSE referenced) before trusting it with network access or credentials. (5) If you want the agent to autonomously execute trades, require explicit confirmations, limit permissions, and monitor network/file activity; otherwise run only backtests with test data first.

FinRL 强化学习交易 (finrl-rl-trading)

Use ensemble deep reinforcement learning (A2C, DDPG, PPO, TD3, SAC) to execute automated multi-market stock tr。

Pipeline

data_collection -> data_storage -> factor_computation -> target_selection -> trading_execution -> visualization

Top Use Cases (14 total)

Ensemble Stock Trading ICAIF 2020 (UC-101)

Executing automated stock trading using an ensemble of multiple DRL agents (A2C, DDPG, PPO, TD3, SAC) to reduce individual agent weakness and improve Triggers: ensemble trading, multiple agents, stock trading

NeurIPS 2018 DRL Training (UC-107)

Training deep reinforcement learning agents (A2C, DDPG, PPO, SAC, TD3) for automated stock trading using the StockTradingEnv environment Triggers: DRL training, stock trading, A2C

NeurIPS 2018 Ensemble Backtesting (UC-108)

Backtesting multiple trained DRL agents against baseline strategies (MVO, DJIA) to evaluate and compare ensemble trading performance Triggers: backtesting, ensemble, DRL agents

For all 14 use cases, see references/USE_CASES.md.

Execute trigger: When user intent matches intent_router.uc_entries[].positive_terms AND user uses action verb (run/execute/跑/执行/backtest/fetch/collect)

What I'll Ask You

• Target market: A-share (default), HK, or crypto? (US stocks in ZVT are half-baked — stockus_nasdaq_AAPL exists but coverage is thin)
• Data source / provider: eastmoney (free, no account), joinquant (account+paid), baostock (free, good history), akshare, or qmt (broker)?
• Strategy type: MACD golden-cross, MA crossover, volume breakout, fundamental screen, or custom factor?
• Time range: start_timestamp and end_timestamp for backtest period
• Target entity IDs: specific stocks (stock_sh_600000) or index components (SZ1000)?

Semantic Locks (Fatal)

ID	Rule	On Violation
SL-01	Execute sell orders before buy orders in every trading cycle	halt
SL-02	Trading signals MUST use next-bar execution (no look-ahead)	halt
SL-03	Entity IDs MUST follow format entity_type_exchange_code	halt
SL-04	DataFrame index MUST be MultiIndex (entity_id, timestamp)	halt
SL-05	TradingSignal MUST have EXACTLY ONE of: position_pct, order_money, order_amount	halt
SL-06	filter_result column semantics: True=BUY, False=SELL, None/NaN=NO ACTION	halt
SL-07	Transformer MUST run BEFORE Accumulator in factor pipeline	halt
SL-08	MACD parameters locked: fast=12, slow=26, signal=9	halt

Full lock definitions: references/LOCKS.md

Top Anti-Patterns (25 total)

• AP-ZVT-183: 除权因子为 inf/NaN 时直接参与乘法导致复权静默失败
• AP-ZVT-179: 第三方数据接口超限后异常被吞噬，数据静默缺失
• AP-ZVT-183B: HFQ（后复权）与 QFQ（前复权）K 线表使用错误导致因子计算漂移

All 25 anti-patterns: references/ANTI_PATTERNS.md

Evidence Quality Notice

[QUALITY NOTICE] This crystal was compiled from blueprint finance-bp-061. Evidence verify ratio = 18.9% and audit fail total = 32. Generated results may have uncaptured requirement gaps. Verify critical decisions against source files (LATEST.yaml / LATEST.jsonl).

Reference Files

File	Contents	When to Load
references/seed.yaml	V6+ 全量权威 (source-of-truth)	有行为/决策争议时必读
references/ANTI_PATTERNS.md	25 条跨项目反模式	开始实现前
references/WISDOM.md	跨项目精华借鉴	架构决策时
references/CONSTRAINTS.md	domain + fatal 约束	规则冲突时
references/USE_CASES.md	全量 KUC-* 业务场景	需要完整示例时
references/LOCKS.md	SL-* + preconditions + hints	生成回测/交易代码前
references/COMPONENTS.md	AST 组件地图（按 module 拆分）	查 API 时

---

Compiled by Doramagic crystal-compilation-v6.1 from finance-bp-061 blueprint at 2026-04-22T13:00:18.884984+00:00. See human_summary.md for non-technical overview.